Being an early adopter of new tech for about three decades now, why wouldn’t I want to jump right into using the new ChatGPT Atlas web browser? Well, that was my thinking before I read a recent article in Digital Trends.
The following paragraphs are excerpts from that article:
ChatGPT can also connect with your Gmail inbox, WhatsApp, Slack, cloud storage services, and more.
With agent mode, you can go shopping or make reservations with just a text prompt. It works autonomously by handling the clicks and types on your behalf. It’s pretty surreal to witness in action, but that’s where the problems begin.
Can you trust an AI agent (and an integrated browser) with sensitive data, such as login credentials? In ChatGPT Atlas, for example, the browser remembers not only your web surfing activity, but also your chat history. And the way it logs all that information is extremely scary. It opens a whole new world of hyper-personalized advertisement and privacy risks.
~ ~ ~
“These tools are contextual engines trained on our behaviors, inputs, and queries. What happens when your browser knows more about you than your partner?” he adds. The risks are grave. In August, user conversations with ChatGPT and xAI’s Grok chatbot were leaked, exposing thousands of conversations on Google Search. Back then, experts pointed out that chatbots are a deeper privacy disaster than browsers.
Mixing the two sure sounds like a calamitous recipe. “If anyone is considering being an early adopter, I’d recommend using these tools cautiously, giving them minimal permissions and blocking their abilities for potentially damaging actions,” warns Feinberg, whose company works at the intersection of AI and security.
An engineer at an AI company, talking on condition of anonymity, told me that despite working on consumer AI products, they are wary of connecting all the services you use with a single chatbot ecosystem. AI assistants are a lot more personal than web browsers, and linking them to other products that you use on a daily basis — from Amazon to Spotify — is like letting an AI company profile your entire life.
One has to draw the line, one where they can balance the convenience of AI browsers with their privacy perils. The shift, however, is inevitable.
(end of excerpts)
For me, I’m happy to wait for what Apple — the reigning King of Privacy — may come up with. I’ll have much more to say then. And I predict it won’t be long. I can even wait till 2026.
Geoff Atkinson, marketing chief at Overstock.com, reiterated that “traffic is just getting so much more expensive to acquire.” His firm started using BT about a year ago. “When a customer first arrives, for example, we log how he or she got there — by a certain keyword, for example. Then, on the next visit, we know that customer and feed them something that’s relevant.” Brent Hieggelke, VP of strategic marketing at Omniture, said when he first joined the firm, his friends thought all this “on-site targeting was hocus-pocus.” But it’s quickly become for real. His firm calls it “automated 1:1 targeting.” The customer hits the site, and their technology builds a profile. “It’s a self-learning predictive modeling engine,” he said. “The optimal content decision can then be sent to the CMS (content management system).” What data is used to select content? “Site behaviors, temporal aspects (such as time of day), environmental aspects, and referrer values,” he said. “This enables companies to quit having those weekly meetings to decide what goes on on the home page.” It now can all be automated, down to the individual. Hieggelke also noted his firm has found that the log-off page is a great place for targeted ads. “Behavioral targeting is bringing marketing back to the marketing world.” A few good questions then came from the audience. The first was “What percentage of users totally wipe out their cookies regularly?” Omniture’s Hieggelke said he’s seen some say as high as 15-30%. “But we find it’s only in the single digits.” A second question related to customers’ concerns for privacy. “We think it’s important to keep a customer mindset,” said Kefta’s Suchet. “Don’t capture too much data — find just what’s relevant to you.” Omniture’s Hieggelke added: “A customer’s name and social security number has no value to us at all.” Suchet added that BT enables campaigns to be “continually learning, changing — you can’t sit still, you must always adjust based on what competitors are doing and so forth.” A final audience question: When building profiles, what data do you use? “Primarily clickstream data,” said Omniture’s Hieggelke, “because that’s easy. But also data from your CRM system, and whatever else is determined to be predictive. Most companies take existing web analystics data and feed that in first.”
institutions that now have it. But a big takeaway is that all this won’t happen soon. Scott Mitic of 
Recent Comments